Is any of the following true about what is needed to achieve this:
- continuous change in its rotation,
- once and for all giving it the right spin to begin with,
- it happens naturally.
The answer is "yes" to all three questions.
If a vehicle is shaped right and is given the right rotation to start with, torques that naturally occur such as gravity gradient torque and torque from atmospheric drag from can help keep the vehicle rotating in the desired orientation. However, this is never perfect and there are always residual undesired torques.
Vehicles need to have some kind of active attitude control system so they can keep themselves properly oriented. If that attitude control relies on fuel, the depletion of the fuel tanks marks the end of the vehicle's useful life.
Update: Approaches to attitude control
The vehicle can only do this so often before it runs out of fuel. For most vehicles, that's the end of the mission. Approaches that reduce the need to use thrusters will extend the vehicle's useful life or enable a bigger payload. In some cases thee alternate approaches entirely eliminate the need for thrusters.
Take advantage of torques from the environment.
Vehicles from Landsat to the Space Station take advantage of rather than fight the external torques exerted on the vehicle by the environment. Environmental torques include gravity gradient torque, atmospheric torque, and magnetic torque. (There's also solar radiation pressure torque, but this is a tiny disturbance.) Some small vehicles in low Earth orbit equipped with magnetic torquers don't use any fuel. They remain functional until they reenter the atmosphere.
Take advantage of rotation.
A rotating object has angular momentum, which makes it harder to turn than if the object wasn't rotating. This adds stability to the vehicle (but also instabilities in some cases). Some of the earliest satellites were spin stabilized.
The next step up in complexity is to construct the vehicle so that it has comprises two parts that rotate about a common axis but at different speeds. Most communications satellites are dual spin satellites. The rotor (plastered with solar arrays) rotates rather quickly for stability while the communications platform rotates but once per day.
Another approach is to place the rotating parts inside the vehicle. These internal rotating devices include momentum wheels, reaction wheels, and control moment gyros. A momentum wheel, like the rotor in a communications satellite, is intended to rotate at a constant angular velocity. A motor with a simple controller is needed to bring the wheel up to speed and then keep it at that speed.
Adding the ability to change the commanded rotation rate to that momentum wheel controller turns the momentum wheel into a reaction wheel. With this ability, angular momentum can be transferred between the main body of the spacecraft to the reaction wheel. A vehicle with three reaction wheels, one per rotation axis, provides an active means of controlling vehicle rotation. Reaction wheels have a basic problem in that rotation speed must be between a minimum value (lest the stabilizing influence be lost) and a maximum value (lest the wheel lose structural integrity). A vehicle that uses reaction wheels needs some alternate control mechanism to help keep the vehicle stable while reaction wheels at their limits are brought back to the nominal rotation rate.
An alternative approach is a control moment gyro (CMG). These are essentially momentum wheels with another motor that pushes against the rotating wheel. (Think of the apocryphal stories of physicists who put airplane gyros in suitcases and then spun them up as a practical joke.) The amount of torque generated by CMGs per unit of power applied can be quite impressive. Just as reaction wheels have operational issues, so do CMGs. In the case of CMGs the problem is gimbal lock. Rotations about one or more axes eventually become uncontrollable. A vehicle that uses CMGs needs some alternate control mechanism to help keep the vehicle stable while CMGs are restored to their nominal rotation axes.